Location: Google Offices, Building F, Zurich.
Date: 12th December 2019
In attendance: Martin Splitt (Google, speaker), John Muller (Google, speaker), Daniel Waisburg (Google, speaker), Dixon Jones (inlinks/Majestic), Aledya Solis (Orainti), Judith Lewis (late), Dawn Anderson (Dawnieando), Juan from Sistrix, and about 150-200 other people.
How do you get invited to one of these events?
Apparently, Google initially wanted to invite 100 people to this event, so Google announced the event I believe in one of their office hangouts. Registration was full in minutes. I don’t know why I was invited vs anyone else… but in short… keep an eye on office hangouts on Youtube with John Muller, or note when this URL changes.
The event was great for Networking: As Martin started, he asked how many people had already spoken with someone they know… EVERYONE put their hands up, and this was just before the start.
These events are going to continue around the world through 2020, so there may be plenty of opportunities to get to one for yourselves
Twitter tag: #WMCZRH
Presentation Notes: Daniel Waisberg on Search Console, behind the scenes.
Daniel claims to be a failed clown, who turned to search in his misery… He is based in Tel Aviv, so presumably, most of the Search Console work is also done there.
The Google Search Console department is made up of several teams:
Search analytics team: Tracks all impressions (many billions of search results daily). 16 months of data. Support slice and dice to help you find insights. 1 guy managed to reduce the time to process this by 10X recently.
Reporting team: Track crawled pages.
Testing tools team: Trigger the entire Google index stack for any given URL. On demand by URL, Hi-fidelity searcg, Support full debugging flow. So ability to test code before production.
Accounts team: Manages verification etc
Alerts team: understand issues over time, they have loads of alert types, makes sure you don’t miss important changes with your site.
How the GSC Team approach their role
When they develop things for Search Console, the consider three steps:
- Define what can help a page succeed on Google. Try to gather all that information in one place
- Classify all the pages in the Google index and annotate issues. (Not necessarily viewable to us)
- Help site owners fix their site.
Daniel went through a recent example. They recently launched the Speed console report. This was very hard because so many teams involved.
Step 1 was to identify metrics that count as “fast”. They decided to use the Chrome UX report as a measurement.
Step 2 used real world data to map the metrics into a usable form. Not easy at scale!
Step 3: produce a report for users. The report can show insights and validate fixes.
What else is new in GSC?
They recently launched a new Message panel (last week). All your messages are strored there. Now you can also filter by category and mark as read/unread. There are also quick links through to the report behind the feature.
My overall takeaway from the talk: Google’s major challenge is being able to cope with scale.
There were lots of Q&A:
Question: Why are some pages marked as noindexed by Google when they shouldn’t be.
Answer: Send Google a message because they need more information. (A feedback form is part of the conference.)
Question: When pages are marked as crawled but not indexed, the documentation says no need to resubmit… but in the questioner’s experience, resubmitting gets them indexed.
Answer: John simply said that they just do not index everything. There are many reasons for not indexing content, usually soft 404s, or duplicate or very similar content. Also, if they drop a URL, it may never get re-seen by Google ever again.
Question from Dawn: What would you say are the main priorities when looking at search.
Answer: They responded that they always try and show the most severe issues first. So work with these first UNLESS there are template issues that let you fix 2000 pages at once.
Presentation Notes: John Muller on changes at Google in 2019
John recapped what has been changing in search at Google over the last year. His team is about a dozen people. Their goal is to help websites to be successful through search… but every site is very different (like cheese, apparently). So they want to give a little bit of something for everyone.
Crawling & rendering news: Goodbye Flash. They no longer care about it. Goodbye, rel=next, rel=prev. Goodbye noindex in robots.txt. But they are making a standard for robots.txt. (Until now, robots.txt had no official standards.)
The Rel=nofollow change is a hint to the future… especiually with the introduction of rel=sponsored & rel=ugc.
They fixed bunch of indexing things that were broken. Rel=canonical, meta robots noidex/nosnippet will remain.
Regarding Rel=next etc: They are working on something clearer, but his thinking is that paginated content comes in two sorts… linking all pages vs drilldown links that go deeper into a side topic detail.
Regarding Mega-menus: He said it just kinda works, but try and use some structure. In short, he didn’t seem to say they were bad.
Crawl Budget: This is their way of preventing server overload. They consider Demand, capacity & necessity..
- To increase demand: make awesome content
- To increase capacity: speed up the site
- To reduce necessity: Site maps etc.
Javascript tips:
- Limit CPU consumption. They Interrupt script.
- Don’t use # in urls for javascript… use real URLs.
Evergreen googlebot basically means crawl is done by chrome. 90th percentile is now rendered within 5 seconds of the initial crawl. (Actually… when Martin spoke later, this was 50th percentile in 5 seconds and the 5 seconds was the QUEUE time, not the RENDERING time. This is something of great interest to me, but I could not get a clear answer on the Renedering time, although Martin did talk in more depth about the Javascript rendering later in the day).
Canonicalization: uses more than rel=canonical (redirects etc).
Mobile first: eventually will be for ALL sites.
m-dot sites are becoming a real problem for them. Redirect users with 301/302 to right version, make signals to Google clear, but probably it is time to rebuild m-dot sites and make your main sites responsive.
Tip: Don’t block mobile version in robots.txt.
Things like loading on interaction (like scroll vs page load) is very hard for Google to test.
Looking forward: They will support MORE STRUCTURED DATA TYPES (be flexible)
Search Results Changes
Less slow sites are showing. AMP has developed (whether you like it or not) but the diverse opinions about it were noted (slightly more hated AMP than liked AMP in a straw poll in the room). He reiterated that there is no AMP ranking boost though (except that in Discover… AMP shows larger images by default).
“BERT”: Goal understand pages and quyeries better, so more relevant pages show.
No penaly or fix BUT you can make better sites.
English only for queries at the moment
Analytics changes: Search Analytics one-box. Droped the Link: and info: searches. Introduced Discover / Google feed to help diversify results.
Crystal Ball: data-nosnippet attribute later this year.
Search Console changes this year:
- URL inspection tool
- 16 months of data is now shown
- Domain verification has developed
- They now have monthly emails (experimental)
- Structured data tools
- Discover reporting
- Amp reporting
Other things this year:
- Lighthouse SEO testing tool (which he described as “pretty cool”.)
- Google+ died.
- EAT started
- sitekit
- They have a wordpress plugin.
Spppeeeed.
Lab tools. Field data (ScUX in Search Console)
Hard to measure.
Images:
- Use alt attributes captions, filnames H1-H3 section headings.
- New preview formats… eg structured data. Use Json-LD. Does MORE than just the rich results.
Q&A Notes from this session:
Hashtags get curtailed in crawls almost always
NoFollows: They are not YET crawling them, but they plan to in the future. But they don’t plan to pass PageRank through, they want to follow them for specific purposes (although they refused to be specific) (Gary)
Q: Are there pans to exted the GSC data beyond 16 months:
A: no. (It was a LOT of work to get to 16 months)
Q: GSC – export data tool to BigQuery or CSV plans?
A: They’ll think about it.
Q: Multilingual sites getting merged into one domain… good or bad?
In the long term, good, but expect fluctuations in the short term. Check Hreflang documentation because it is hard to implement.
Q: In GSC do you plan to add separate data for voice search?
A: No. There is the issue “what is voice data?” and what “would you do with that data?” … so sounds a way off… do whatever you are doing for featured snippets. Really.
Q: GSC API plans?
A: They know there are a few people that want it, but not enough to prioritize this work. Only expect APIs for things that have been stable for a really long time, because changing an API causes problems for customers.
Q: Plans to report for Google Discover?
A: It is complete in GSC. (So no plans to expand into GA.) You can use the URLs in Discover to help surface topics that you should be addressing.
Q: Where .JS is being crawled that is being used for (user) tracking, this is casuing problems for one person, creating weird warnings and issues in GSC.
A: You can block specific JS in Robots.txt (Daniel).
Q: Geotargeting content by IP – is this OK?
A: Googlebot only crawls from the US… so don’t deliver content ONLY by IP address of the user, if you want Googlebot to see the non-us version of the content.
0 Comments